unet/diffusion.py

"""HSIGene diffusion modules - UNet, ResBlock, etc. From openaimodel."""

from abc import abstractmethod
import math

import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F

from .utils import (
    checkpoint,
    conv_nd,
    linear,
    zero_module,
    normalization,
    timestep_embedding,
    exists,
)
from .attention import SpatialTransformer


def avg_pool_nd(dims, *args, **kwargs):
    """Create a 1D, 2D, or 3D average pooling module."""
    if dims == 1:
        return nn.AvgPool1d(*args, **kwargs)
    elif dims == 2:
        return nn.AvgPool2d(*args, **kwargs)
    elif dims == 3:
        return nn.AvgPool3d(*args, **kwargs)
    raise ValueError(f"unsupported dimensions: {dims}")


def convert_module_to_f16(x):
    pass


def convert_module_to_f32(x):
    pass


class TimestepBlock(nn.Module):
    """Any module where forward() takes timestep embeddings as a second argument."""

    @abstractmethod
    def forward(self, x, emb):
        """Apply the module to `x` given `emb` timestep embeddings."""


class TimestepEmbedSequential(nn.Sequential, TimestepBlock):
    """Sequential module that passes timestep embeddings to children that support it."""

    def forward(self, x, emb, context=None):
        for layer in self:
            if isinstance(layer, TimestepBlock):
                x = layer(x, emb)
            elif isinstance(layer, SpatialTransformer):
                x = layer(x, context)
            else:
                x = layer(x)
        return x


class Upsample(nn.Module):
    """Upsampling layer with optional convolution."""

    def __init__(self, channels, use_conv, dims=2, out_channels=None, padding=1):
        super().__init__()
        self.channels = channels
        self.out_channels = out_channels or channels
        self.use_conv = use_conv
        self.dims = dims
        if use_conv:
            self.conv = conv_nd(dims, self.channels, self.out_channels, 3, padding=padding)

    def forward(self, x):
        assert x.shape[1] == self.channels
        if self.dims == 3:
            x = F.interpolate(
                x, (x.shape[2], x.shape[3] * 2, x.shape[4] * 2), mode="nearest"
            )
        else:
            x = F.interpolate(x, scale_factor=2, mode="nearest")
        if self.use_conv:
            x = self.conv(x)
        return x


class Downsample(nn.Module):
    """Downsampling layer with optional convolution."""

    def __init__(self, channels, use_conv, dims=2, out_channels=None, padding=1):
        super().__init__()
        self.channels = channels
        self.out_channels = out_channels or channels
        self.use_conv = use_conv
        self.dims = dims
        stride = 2 if dims != 3 else (1, 2, 2)
        if use_conv:
            self.op = conv_nd(
                dims, self.channels, self.out_channels, 3, stride=stride, padding=padding
            )
        else:
            assert self.channels == self.out_channels
            self.op = avg_pool_nd(dims, kernel_size=stride, stride=stride)

    def forward(self, x):
        assert x.shape[1] == self.channels
        return self.op(x)


class ResBlock(TimestepBlock):
    """Residual block with timestep conditioning."""

    def __init__(
        self,
        channels,
        emb_channels,
        dropout,
        out_channels=None,
        use_conv=False,
        use_scale_shift_norm=False,
        dims=2,
        use_checkpoint=False,
        up=False,
        down=False,
    ):
        super().__init__()
        self.channels = channels
        self.emb_channels = emb_channels
        self.dropout = dropout
        self.out_channels = out_channels or channels
        self.use_conv = use_conv
        self.use_checkpoint = use_checkpoint
        self.use_scale_shift_norm = use_scale_shift_norm

        self.in_layers = nn.Sequential(
            normalization(channels),
            nn.SiLU(),
            conv_nd(dims, channels, self.out_channels, 3, padding=1),
        )

        self.updown = up or down
        if up:
            self.h_upd = Upsample(channels, False, dims)
            self.x_upd = Upsample(channels, False, dims)
        elif down:
            self.h_upd = Downsample(channels, False, dims)
            self.x_upd = Downsample(channels, False, dims)
        else:
            self.h_upd = self.x_upd = nn.Identity()

        self.emb_layers = nn.Sequential(
            nn.SiLU(),
            linear(
                emb_channels,
                2 * self.out_channels if use_scale_shift_norm else self.out_channels,
            ),
        )
        self.out_layers = nn.Sequential(
            normalization(self.out_channels),
            nn.SiLU(),
            nn.Dropout(p=dropout),
            zero_module(conv_nd(dims, self.out_channels, self.out_channels, 3, padding=1)),
        )

        if self.out_channels == channels:
            self.skip_connection = nn.Identity()
        elif use_conv:
            self.skip_connection = conv_nd(dims, channels, self.out_channels, 3, padding=1)
        else:
            self.skip_connection = conv_nd(dims, channels, self.out_channels, 1)

    def forward(self, x, emb):
        return checkpoint(self._forward, (x, emb), self.parameters(), self.use_checkpoint)

    def _forward(self, x, emb):
        if self.updown:
            in_rest, in_conv = self.in_layers[:-1], self.in_layers[-1]
            h = in_rest(x)
            h = self.h_upd(h)
            x = self.x_upd(x)
            h = in_conv(h)
        else:
            h = self.in_layers(x)
        emb_out = self.emb_layers(emb).type(h.dtype)
        while len(emb_out.shape) < len(h.shape):
            emb_out = emb_out[..., None]
        if self.use_scale_shift_norm:
            out_norm, out_rest = self.out_layers[0], self.out_layers[1:]
            scale, shift = emb_out.chunk(2, dim=1)
            h = out_norm(h) * (1 + scale) + shift
            h = out_rest(h)
        else:
            h = h + emb_out
            h = self.out_layers(h)
        return self.skip_connection(x) + h


class AttentionBlock(nn.Module):
    """Spatial self-attention block."""

    def __init__(
        self,
        channels,
        num_heads=1,
        num_head_channels=-1,
        use_checkpoint=False,
        use_new_attention_order=False,
    ):
        super().__init__()
        self.channels = channels
        if num_head_channels == -1:
            self.num_heads = num_heads
        else:
            assert channels % num_head_channels == 0
            self.num_heads = channels // num_head_channels
        self.use_checkpoint = use_checkpoint
        self.norm = normalization(channels)
        self.qkv = conv_nd(1, channels, channels * 3, 1)
        self.attention = (
            QKVAttention(self.num_heads)
            if use_new_attention_order
            else QKVAttentionLegacy(self.num_heads)
        )
        self.proj_out = zero_module(conv_nd(1, channels, channels, 1))

    def forward(self, x):
        return checkpoint(self._forward, (x,), self.parameters(), True)

    def _forward(self, x):
        b, c, *spatial = x.shape
        x = x.reshape(b, c, -1)
        qkv = self.qkv(self.norm(x))
        h = self.attention(qkv)
        h = self.proj_out(h)
        return (x + h).reshape(b, c, *spatial)


class QKVAttentionLegacy(nn.Module):
    """QKV attention - split heads before split qkv."""

    def __init__(self, n_heads):
        super().__init__()
        self.n_heads = n_heads

    def forward(self, qkv):
        bs, width, length = qkv.shape
        assert width % (3 * self.n_heads) == 0
        ch = width // (3 * self.n_heads)
        q, k, v = qkv.reshape(bs * self.n_heads, ch * 3, length).split(ch, dim=1)
        scale = 1 / math.sqrt(math.sqrt(ch))
        weight = torch.einsum("bct,bcs->bts", q * scale, k * scale)
        weight = torch.softmax(weight.float(), dim=-1).type(weight.dtype)
        a = torch.einsum("bts,bcs->bct", weight, v)
        return a.reshape(bs, -1, length)


class QKVAttention(nn.Module):
    """QKV attention - split qkv before split heads."""

    def __init__(self, n_heads):
        super().__init__()
        self.n_heads = n_heads

    def forward(self, qkv):
        bs, width, length = qkv.shape
        assert width % (3 * self.n_heads) == 0
        ch = width // (3 * self.n_heads)
        q, k, v = qkv.chunk(3, dim=1)
        scale = 1 / math.sqrt(math.sqrt(ch))
        weight = torch.einsum(
            "bct,bcs->bts",
            (q * scale).view(bs * self.n_heads, ch, length),
            (k * scale).view(bs * self.n_heads, ch, length),
        )
        weight = torch.softmax(weight.float(), dim=-1).type(weight.dtype)
        a = torch.einsum("bts,bcs->bct", weight, v.reshape(bs * self.n_heads, ch, length))
        return a.reshape(bs, -1, length)


class UNetModel(nn.Module):
    """Full UNet with attention and timestep embedding."""

    def __init__(
        self,
        image_size,
        in_channels,
        model_channels,
        out_channels,
        num_res_blocks,
        attention_resolutions,
        dropout=0,
        channel_mult=(1, 2, 4, 8),
        conv_resample=True,
        dims=2,
        num_classes=None,
        use_checkpoint=False,
        use_fp16=False,
        num_heads=-1,
        num_head_channels=-1,
        num_heads_upsample=-1,
        use_scale_shift_norm=False,
        resblock_updown=False,
        use_new_attention_order=False,
        use_spatial_transformer=False,
        transformer_depth=1,
        context_dim=None,
        n_embed=None,
        legacy=True,
        disable_self_attentions=None,
        num_attention_blocks=None,
        disable_middle_self_attn=False,
        use_linear_in_transformer=False,
    ):
        super().__init__()
        if use_spatial_transformer:
            assert context_dim is not None
        if context_dim is not None:
            assert use_spatial_transformer
            if hasattr(context_dim, "__iter__") and not isinstance(context_dim, (list, tuple)):
                context_dim = list(context_dim)

        if num_heads_upsample == -1:
            num_heads_upsample = num_heads
        if num_heads == -1:
            assert num_head_channels != -1
        if num_head_channels == -1:
            assert num_heads != -1

        self.image_size = image_size
        self.in_channels = in_channels
        self.model_channels = model_channels
        self.out_channels = out_channels
        if isinstance(num_res_blocks, int):
            self.num_res_blocks = len(channel_mult) * [num_res_blocks]
        else:
            assert len(num_res_blocks) == len(channel_mult)
            self.num_res_blocks = num_res_blocks

        self.attention_resolutions = attention_resolutions
        self.dropout = dropout
        self.channel_mult = channel_mult
        self.conv_resample = conv_resample
        self.num_classes = num_classes
        self.use_checkpoint = use_checkpoint
        self.dtype = torch.float16 if use_fp16 else torch.float32
        self.num_heads = num_heads
        self.num_head_channels = num_head_channels
        self.num_heads_upsample = num_heads_upsample
        self.predict_codebook_ids = n_embed is not None

        time_embed_dim = model_channels * 4
        self.time_embed = nn.Sequential(
            linear(model_channels, time_embed_dim),
            nn.SiLU(),
            linear(time_embed_dim, time_embed_dim),
        )

        if num_classes is not None:
            if isinstance(num_classes, int):
                self.label_emb = nn.Embedding(num_classes, time_embed_dim)
            elif num_classes == "continuous":
                self.label_emb = nn.Linear(1, time_embed_dim)
            else:
                raise ValueError()

        self.input_blocks = nn.ModuleList(
            [TimestepEmbedSequential(conv_nd(dims, in_channels, model_channels, 3, padding=1))]
        )
        self._feature_size = model_channels
        input_block_chans = [model_channels]
        ch = model_channels
        ds = 1

        for level, mult in enumerate(channel_mult):
            for nr in range(self.num_res_blocks[level]):
                layers = [
                    ResBlock(
                        ch,
                        time_embed_dim,
                        dropout,
                        out_channels=mult * model_channels,
                        dims=dims,
                        use_checkpoint=use_checkpoint,
                        use_scale_shift_norm=use_scale_shift_norm,
                    )
                ]
                ch = mult * model_channels
                if ds in attention_resolutions:
                    if num_head_channels == -1:
                        dim_head = ch // num_heads
                    else:
                        num_heads_cur = ch // num_head_channels
                        dim_head = num_head_channels
                    if legacy:
                        dim_head = ch // num_heads if use_spatial_transformer else num_head_channels
                    disabled_sa = (
                        disable_self_attentions[level]
                        if exists(disable_self_attentions)
                        else False
                    )
                    if not exists(num_attention_blocks) or nr < num_attention_blocks[level]:
                        attn_block = (
                            AttentionBlock(
                                ch,
                                use_checkpoint=use_checkpoint,
                                num_heads=num_heads,
                                num_head_channels=dim_head,
                                use_new_attention_order=use_new_attention_order,
                            )
                            if not use_spatial_transformer
                            else SpatialTransformer(
                                ch,
                                num_heads,
                                dim_head,
                                depth=transformer_depth,
                                context_dim=context_dim,
                                disable_self_attn=disabled_sa,
                                use_linear=use_linear_in_transformer,
                                use_checkpoint=use_checkpoint,
                            )
                        )
                        layers.append(attn_block)
                self.input_blocks.append(TimestepEmbedSequential(*layers))
                self._feature_size += ch
                input_block_chans.append(ch)
            if level != len(channel_mult) - 1:
                out_ch = ch
                down_block = (
                    ResBlock(
                        ch,
                        time_embed_dim,
                        dropout,
                        out_channels=out_ch,
                        dims=dims,
                        use_checkpoint=use_checkpoint,
                        use_scale_shift_norm=use_scale_shift_norm,
                        down=True,
                    )
                    if resblock_updown
                    else Downsample(ch, conv_resample, dims=dims, out_channels=out_ch)
                )
                self.input_blocks.append(TimestepEmbedSequential(down_block))
                ch = out_ch
                input_block_chans.append(ch)
                ds *= 2
                self._feature_size += ch

        if num_head_channels == -1:
            dim_head = ch // num_heads
        else:
            num_heads_cur = ch // num_head_channels
            dim_head = num_head_channels
        if legacy:
            dim_head = ch // num_heads if use_spatial_transformer else num_head_channels
        mid_attn = (
            AttentionBlock(
                ch,
                use_checkpoint=use_checkpoint,
                num_heads=num_heads,
                num_head_channels=dim_head,
                use_new_attention_order=use_new_attention_order,
            )
            if not use_spatial_transformer
            else SpatialTransformer(
                ch,
                num_heads,
                dim_head,
                depth=transformer_depth,
                context_dim=context_dim,
                disable_self_attn=disable_middle_self_attn,
                use_linear=use_linear_in_transformer,
                use_checkpoint=use_checkpoint,
            )
        )
        self.middle_block = TimestepEmbedSequential(
            ResBlock(
                ch,
                time_embed_dim,
                dropout,
                dims=dims,
                use_checkpoint=use_checkpoint,
                use_scale_shift_norm=use_scale_shift_norm,
            ),
            mid_attn,
            ResBlock(
                ch,
                time_embed_dim,
                dropout,
                dims=dims,
                use_checkpoint=use_checkpoint,
                use_scale_shift_norm=use_scale_shift_norm,
            ),
        )
        self._feature_size += ch

        self.output_blocks = nn.ModuleList([])
        for level, mult in list(enumerate(channel_mult))[::-1]:
            for i in range(self.num_res_blocks[level] + 1):
                ich = input_block_chans.pop()
                layers = [
                    ResBlock(
                        ch + ich,
                        time_embed_dim,
                        dropout,
                        out_channels=model_channels * mult,
                        dims=dims,
                        use_checkpoint=use_checkpoint,
                        use_scale_shift_norm=use_scale_shift_norm,
                    )
                ]
                ch = model_channels * mult
                if ds in attention_resolutions:
                    if num_head_channels == -1:
                        dim_head = ch // num_heads
                    else:
                        num_heads_cur = ch // num_head_channels
                        dim_head = num_head_channels
                    if legacy:
                        dim_head = (
                            ch // num_heads if use_spatial_transformer else num_head_channels
                        )
                    disabled_sa = (
                        disable_self_attentions[level]
                        if exists(disable_self_attentions)
                        else False
                    )
                    if not exists(num_attention_blocks) or i < num_attention_blocks[level]:
                        attn_block = (
                            AttentionBlock(
                                ch,
                                use_checkpoint=use_checkpoint,
                                num_heads=num_heads_upsample,
                                num_head_channels=dim_head,
                                use_new_attention_order=use_new_attention_order,
                            )
                            if not use_spatial_transformer
                            else SpatialTransformer(
                                ch,
                                num_heads,
                                dim_head,
                                depth=transformer_depth,
                                context_dim=context_dim,
                                disable_self_attn=disabled_sa,
                                use_linear=use_linear_in_transformer,
                                use_checkpoint=use_checkpoint,
                            )
                        )
                        layers.append(attn_block)
                if level and i == self.num_res_blocks[level]:
                    out_ch = ch
                    up_block = (
                        ResBlock(
                            ch,
                            time_embed_dim,
                            dropout,
                            out_channels=out_ch,
                            dims=dims,
                            use_checkpoint=use_checkpoint,
                            use_scale_shift_norm=use_scale_shift_norm,
                            up=True,
                        )
                        if resblock_updown
                        else Upsample(ch, conv_resample, dims=dims, out_channels=out_ch)
                    )
                    layers.append(up_block)
                    ds //= 2
                self.output_blocks.append(TimestepEmbedSequential(*layers))
                self._feature_size += ch

        self.out = nn.Sequential(
            normalization(ch),
            nn.SiLU(),
            zero_module(conv_nd(dims, model_channels, out_channels, 3, padding=1)),
        )
        if self.predict_codebook_ids:
            self.id_predictor = nn.Sequential(
                normalization(ch),
                conv_nd(dims, model_channels, n_embed, 1),
            )

    def forward(self, x, timesteps=None, metadata=None, context=None, y=None, **kwargs):
        assert (y is not None) == (self.num_classes is not None)
        hs = []
        t_emb = timestep_embedding(timesteps, self.model_channels, repeat_only=False)
        emb = self.time_embed(t_emb)
        if metadata is not None:
            if isinstance(metadata, (list, tuple)) and len(metadata) == 1:
                metadata = metadata[0]
            emb = emb + metadata

        if self.num_classes is not None:
            assert y.shape[0] == x.shape[0]
            emb = emb + self.label_emb(y)

        h = x.type(self.dtype)
        for module in self.input_blocks:
            h = module(h, emb, context)
            hs.append(h)
        h = self.middle_block(h, emb, context)
        for module in self.output_blocks:
            h = torch.cat([h, hs.pop()], dim=1)
            h = module(h, emb, context)
        h = h.type(x.dtype)
        if self.predict_codebook_ids:
            return self.id_predictor(h)
        return self.out(h)